---
title: ChatGPT for venture capital: decks, NDAs and LP data
description: Using ChatGPT at a VC firm: which plans train on pitch decks, what NDAs and side letters say about AI, and how to answer an LP's AI question in a DDQ.
canonical: https://privatesuperintelligence.si/blog/chatgpt-for-venture-capital/
last-updated: 2026-10-15
---

# ChatGPT for venture capital: decks, NDAs and LP data

ChatGPT is useful at a venture firm for research, first drafts and summaries, and the plan you use decides what happens to the founder and LP material you give it. OpenAI says ChatGPT Business, Enterprise and Edu do not train on workspace content by default, while Free, Go, Plus and Pro may train on chats unless you turn that off. Before a pitch deck or a side letter goes in, check two more things: what your NDA says about third-party processing, and how you'll describe the tool when an LP asks.

## Which ChatGPT plans train on a pitch deck

OpenAI's help center and enterprise privacy page set out the defaults. The summary below covers training. Our guide [Is ChatGPT safe for confidential information?](/blog/is-chatgpt-safe-for-confidential-information/) also covers retention and staff access plan by plan.

| Plan | Trains on your content by default? | Notes for a VC firm |
|---|---|---|
| Free, Go, Plus, Pro (personal) | Yes, unless you turn off "Improve the model for everyone" | Check partners' personal accounts, which sit on these plans. Thumbs up or down feedback can send the whole conversation for training, even after opting out |
| Temporary Chat | No | OpenAI may keep a copy for up to 30 days for safety purposes |
| Business | No | OpenAI says it does not train on workspace data. Admins and owners cannot see all private member chats by default |
| Enterprise and Edu | No | Admins control retention. Deleted conversations leave OpenAI's systems within 30 days unless law requires otherwise |
| API | No | Abuse monitoring logs are kept up to 30 days by default. Zero data retention needs OpenAI's approval |

OpenAI says Enterprise, Edu, Business and the API Platform have completed a SOC 2 Type 2 audit, and that it encrypts data at rest with AES-256 and in transit with TLS 1.2 or higher.

Training is one risk among several. In May 2025 a federal court in the New York Times copyright case ordered OpenAI to preserve ChatGPT output logs that would otherwise have been deleted, an obligation that ended on September 26, 2025. In January 2026 Judge Sidney Stein affirmed an order to produce a sample of 20 million de-identified chat logs. Enterprise, Edu and zero data retention API traffic fell outside the original order. A deletion promise can be paused by a court in a case you aren't part of.

## Can I put a pitch deck in ChatGPT?

On a personal plan with default settings, the deck may be used for training. On a business plan it won't be, but it still sits with OpenAI under OpenAI's terms. Whether that's acceptable depends on what you agreed with the founder.

DarrowEverett, a law firm, argued in June 2026 that most NDAs restrict disclosure to "third parties" or limit use to a defined "purpose," and were "written with human recipients in mind." Its view: "Even if an NDA is silent on AI, this act may constitute a breach," because the information goes to infrastructure a technology company runs, under terms the disclosing party never agreed to.

Rules you can adopt without new contracts:

- Keep founder decks, models and cap tables out of personal AI accounts.
- Use a business workspace your firm has reviewed for any deal material.
- Strip company names and figures when you only need help with structure or wording.
- If a founder's NDA restricts disclosure, ask before you use any AI tool on their documents.

This is general information, not legal advice. Have counsel read your own NDA and fund documents.

## What NDAs and side letters say about AI

The same DarrowEverett piece lists four kinds of AI provisions now appearing in NDAs:

| Provision | What it does |
|---|---|
| Express prohibition | Bars putting confidential information into any AI tool, whether or not the tool trains on inputs |
| Expanded definition of disclosure | Counts submission to automated systems, machine learning models or AI platforms as disclosure |
| Permitted-use carve-out | Allows enterprise-grade tools that keep data isolated, disable training on inputs and meet named security standards |
| Internal AI policy representation | Requires the receiving party to confirm it has controls on employee use of AI with confidential information |

If you send founders your own NDA, or you're redrafting your template, the carve-out is the version most compatible with using AI at all. It tells the founder which class of tool you'll use, and it gives your team a clear line.

Side letters work the same way. Read the confidentiality clauses in each LPA and side letter and ask whether sending LP information to an AI provider counts as disclosure to a third party under them. If you can't tell, treat LP names, commitments and capital account data as off limits for any tool your counsel hasn't cleared.

## How to answer an LP's AI question in a DDQ

ILPA's Due Diligence Questionnaire 2.0, released in November 2021, does not contain a question that names AI. It still reaches your AI tools in two places:

- **Section 18 (Data Security, Technology, Third Parties)** asks whether firm data provided to or managed by third-party service providers gets security reviews and independent assessments (18.14), and whether sensitive data is encrypted at rest and in transit (18.15).
- **Appendix D** asks you to list the technology tool used for each function, including deal sourcing, due diligence, portfolio monitoring and reporting and investor relations, and whether a third party performs it.

LPs also have AI-specific questions ready. ILPA's Responsible AI Quick Guide for Asset Owners (January 2024) suggests LPs ask how far a GP relies on AI to make investment decisions, and "What are the GP's controls around AI use at the fund level?"

A good answer names the tool and the plan, says whether it trains on firm data, says who at the provider can access content, and describes your internal rule on founder and LP data. If partners still use personal accounts, fix that before the questionnaire arrives.

## Good uses of ChatGPT at a VC firm, by data risk

| Lower risk | Needs a reviewed business plan | Keep out unless counsel clears it |
|---|---|---|
| Market maps from public sources | Summaries of your own call notes | A founder's data room shared under a strict NDA |
| Drafting a thesis post or job description | First-draft IC memos from approved materials | LP names, commitments and capital accounts |
| Explaining a public company's filing | Portfolio KPI roll-ups from founder updates | Side letter terms |
| Editing your own writing | Replies to founders and co-investors | Wire instructions and LP banking details |

For diligence workflows, read [AI due diligence for venture capital](/blog/ai-due-diligence-for-vc/). For the bigger picture of where firms use AI, read [How VC firms use AI](/blog/how-vc-firms-use-ai/).

## Private SuperIntelligence for venture firms

Private SuperIntelligence is the most capable AI, trusted with everything a venture firm holds: decks, data rooms, the CRM and LP records. Each deployment is built for one firm, works across its own email, documents and data, and answers with citations. The provider keeps no copy of the work and trains nothing on it, which gives you a clear answer for founders and for an LP's DDQ. See [AI for venture capital firms](/venture-capital/).

## Frequently asked questions

### Does ChatGPT train on pitch decks I upload?

On Free, Go, Plus and Pro it may, unless you turn off "Improve the model for everyone." OpenAI says Business, Enterprise and Edu do not train on workspace content by default.

### Is ChatGPT Enterprise safe for VC deal data?

OpenAI says Enterprise doesn't train on your data by default, lets admins control retention and limits employee access to incidents, permitted recovery and legal requirements. Whether that satisfies a given founder NDA or side letter is a contract question for your counsel.

### Does sharing an NDA'd deck with an AI tool breach the NDA?

It may. A June 2026 DarrowEverett analysis argues that sending deal information to an AI provider can count as disclosure to a third party even when the NDA says nothing about AI. This is general information, not legal advice.

### Does the ILPA DDQ ask about AI?

The ILPA DDQ 2.0 has no question that names AI, but it asks about security reviews of data held by third parties and asks you to list the tool used for each function. ILPA's 2024 Responsible AI Quick Guide suggests LPs ask about a GP's controls on AI use.

### What ChatGPT prompts work for venture capital?

Prompts that start from public information or your own notes carry the least risk: market maps, thesis drafts, summaries of call notes and edits to your writing. Ask it to cite where each figure came from so you can check it.


## Related reading

- [How VC firms use AI, and what stays private](/blog/how-vc-firms-use-ai/)
- [AI due diligence for venture capital](/blog/ai-due-diligence-for-vc/)
- [AI for LP reporting and investor relations](/blog/ai-for-lp-reporting/)
- [AI tools for venture capital, and what each sees](/blog/best-ai-tools-for-venture-capital/)
- [Is ChatGPT safe for confidential information?](/blog/is-chatgpt-safe-for-confidential-information/)
- [What is zero data retention AI?](/blog/zero-data-retention-ai/)
- [What is private superintelligence?](/blog/what-is-private-superintelligence/)

## Sources

- [Data controls in ChatGPT (OpenAI Help Center)](https://help.openai.com/en/articles/7730893-data-controls-faq), accessed 2026-10-15
- [ChatGPT Business: General FAQ (OpenAI Help Center)](https://help.openai.com/en/articles/8542115-chatgpt-business-faq), accessed 2026-10-15
- [Enterprise privacy at OpenAI (OpenAI)](https://openai.com/enterprise-privacy/), accessed 2026-10-15
- [Data controls in the OpenAI platform (OpenAI)](https://developers.openai.com/api/docs/guides/your-data), accessed 2026-10-15
- [Is ChatGPT safe for confidential information? (Private SuperIntelligence, October 2026)](/blog/is-chatgpt-safe-for-confidential-information/)
- [How we're responding to The New York Times' data demands (OpenAI)](https://openai.com/index/response-to-nyt-data-demands/), accessed 2026-10-15
- [ChatGPT creator must turn over 20M chat logs in copyright litigation, federal judge says (ABA Journal, January 8, 2026)](https://www.abajournal.com/news/article/chatgpt-creator-must-turn-over-20m-chat-logs-in-copyright-litigation-federal-judge-says)
- [Loose Lips, Large Language Models: What Your NDA is Missing in the Age of AI (DarrowEverett LLP, June 2, 2026)](https://darroweverett.com/ndas-non-disclosure-agreements-age-ai-practical-steps-analysis/)
- [Due Diligence Questionnaire 2.0 (ILPA, November 2021)](https://ilpa.org/wp-content/uploads/2021/11/ILPA-DDQ-2.0.pdf)
- [Responsible AI Quick Guide for Asset Owners (ILPA, January 2024)](https://ilpa.org/wp-content/uploads/2023/10/Responsible-AI-Quick-Guide-for-Asset-Owners.pdf)